Primary Colors: A Mixed Blessing for Al Gore

Primary Colors: A Mixed Blessing for Al GoreAuthor(s): Helmut NorpothSource: PS: Political Science and Politics, Vol. 34, No. 1 (Mar., 2001), pp. 45-48Published by: American Political Science AssociationStable URL: http://www.jstor.org/stable/1350309 .

Accessed: 15/06/2014 09:26

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

American Political Science Association is collaborating with JSTOR to digitize, preserve and extend access toPS: Political Science and Politics.

http://www.jstor.org

This content downloaded from 195.34.79.223 on Sun, 15 Jun 2014 09:26:02 AMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=apsa

http://www.jstor.org/stable/1350309?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


Primary Colors: A Mixed Blessing for AI Gore*

by Helmut Norpoth, SUNY, Stony Brook

Having nailed down his party's presidential nomination with an early and decisive victory in the primaries, Al Gore will go on to defeat George W Bush in the general election by a comfortable margin. The Democratic ticket, so the prediction, will take 55% of the major-party popular vote in November, compared to 45% for the Republican one-almost exactly the reverse of what national polls have been showing as late as June of the election year.

- A forecast issued July 26, 2000

"Prophecy is good business, but full of risks," warned Mark Twain. How true in this election. Even though Al Gore edged George W. Bush by half-a-million votes in the popular count, he certainly did not win by the "comfortable margin" I predicted more than two months before the event. What was the basis of my prediction? And

what may have cut it short?

I based my forecast of a Gore victory on the results from a model whose key predictor is performance in presidential primaries (Norpoth 2000). I estimated the

parameters of this model with data on presidential elections going back as far as 1912, when the first presidential primaries were held. The model's results matched the actual out-comes for all elections held between 1912 and 1996 except in 1960. While the overall record of the model is reassuring, the 1960 miss sounded an ominous warning for 2000.

Unlike almost all other forecasting models (see Campbell and Garand 2000; Lewis-Beck and Rice 1992), mine relies on neither measures of economic performance nor of presidential approval as predictors. Its key predictor, the primary performance of the incumbent-party candidate, does, however, correlate strongly with presidential approval in election years (.78) and, although somewhat less strongly, with economic conditions. Hence, strong primary performance can be seen as a reflection of broad-based satisfaction with government performance. For the incumbent-

party candidate, typically the White House occupant himself, to ride a resounding victory in the primaries to nomination at the national convention is a sure sign that the electorate is pleased with the state of the nation in the election year. In contrast, a poor showing by the incumbent-party candidate in the primaries signals general discontent with the nation's affairs (Atkeson 1998). While the approval rating of the White House occupant would be a more perfect measure of public satisfaction with government performance, it is available in election years only since 1948. Primary performance goes back much further, putting model-building on a firmer footing.

Presidential Primaries The introduction of presidential

primaries in 1912 afforded aspirants for the White House a novel opportunity to gain popular recognition as well as convention delegates.1 Right from their inception, primary elections established a pattern. That year Theodore Roosevelt used the 1912 primaries to challenge the sitting president, Taft, for the nomination of the Republican Party. Although Roosevelt won the primaries, he failed to be rewarded with the party's nomination at its national convention (see Table 1). In November, Taft, the party's but not the people's nominee, lost the general election. Denying the primary winner its nomination, so it seemed, proved costly for the incumbent party. Four years later, with the White House in Democratic hands, Woodrow Wilson won practically unanimous support during his party's primaries and was (re)nominated at the national convention. Wilson, of course, went on to victory in the general election. The lessons from those two elections held throughout the century: A party nominating its primary winner has a good chance to win the November election, whereas a party nominating someone else risks defeat in November.

It stands to reason that a sitting president eyeing a second term should expect to receive strong support for his nomination. To be seriously challenged in the primaries, let alone defeated as

PSOnline www.apsanet.org 45



Taft was is a sure sign of trouble for the general election. Just winning the primaries, however, is not good enough for incumbent presidents. They must do so in a landslide, though not necessarily of Wilsonian proportion. The threshold I adopted for my analysis is 60%, a level typically considered a "landslide." Only a vote share above that level in the primaries is scored positive for incumbent presidents. In contests without an incumbent president in the race, on the other hand, no such qualification applies. In a crowded field, winning simply means getting the most support, even if it is as low as 12%. With the number of candidates in primary contests varying greatly from election to election, it makes little sense to define primary performance by the exact percentage of support. Instead, I apply a simple digital definition:

Positive Presidential candidate is the primary winner (by more than 60% in case of incumbent presidents; New Hampshire primary since 1952, all primaries before 1952).

Negative Presidential candidate is not the primary winner (or the winner of less than 60% in case of incumbent presidents; New Hampshire primary since 1952, all primaries before 1952).

Measured this way, the primary performance of the incumbent-party candidate permitted me to correctly (although retrospectively) predict the winner of all but

one presidential election held between 1912 and 1996. The only miss is the 1960 election, the closest in the twentieth century. Reflecting the positive primary performance by the incumbent-party candidate, Nixon gets the winning nod that year. The primary performance of the out-party candidate lags somewhat in predictive power, although it still indicates who would win 18 of the 22 elections. In two elections (1928 and 1988), prospective winners went down to defeat, and prospective losers triumphed in two other contests (1920 and 1992). It is worth noting that in all these four instances, consideration of the primary performance of the incumbent party leads to picking the right winner. It is only in 1960 that out-party beats in-party performance as a predictor of the November outcome.

The Forecast Model

Aside from primary success, the forecast model also relies on the effects of past elections. The sequence of presidential elections in the U.S. is not white noise; every election is not an entirely new event, utterly independent of any that came before. Tapping that dynamic is the key to the forecaster's art.

As longtime readers of PS will already know (Midlarsky 1984; Norpoth 1995), the presidential vote follows a second-order autoregressive process in a manner suggesting a cycle. The vote division in any given presidential election is positively correlated with

TABLE I Presidential Primary Winners and Convention Nominations, 1912-2000 Year Incumbent Party % Nominated? Opposition Party % Nominated?

1912 T. Roosevelt (R) 51.5 No Wilson (D) 44.6 Yes 1916 Wilson* (D) 98.8 Yes Brumbaugh (R) 12.1 No 1920 Palmer (D) 16.0 No Johnson (R) 30.3 No 1924 Coolidge* (R) 68.4 Yes McAdoo (D) 59.8 No 1928 Hoover (R) 49.2 Yes Smith (D) 39.5 Yes 1932 France (R) 48.5 No F. Roosevelt (D) 44.5 Yes 1936 F. Roosevelt* (D) 92.9 Yes Borah (R) 44.4 No 1940 F. Roosevelt* (D) 71.7 Yes Dewey (R) 49.7 No 1944 F. Roosevelt* (D) 70.9 Yes MacArthur (R) 29.1 No 1948 Truman* (D) 63.9 Yes Warren (R) 29.1 No 1952 Kefauver (D) 55.0 No Eisenhower (R) 50.4 Yes 1956 Eisenhower* (R) 98.9 Yes Kefauver (D) 84.6 No 1960 Nixon (R) 89.3 Yes Kennedy (D) 85.2 Yes 1964 Johnson* (D) 95.3 Yes Lodge (R) 35.5 No 1968 Johnson* (D) 49.6 No Nixon (R) 77.6 Yes 1972 Nixon* (R) 67.6 Yes Muskie (D) 46.4 No 1976 Ford* (R) 49.4 Yes Carter (D) 28.4 Yes 1980 Carter* (D) 47.1 Yes Reagan (R) 49.6 Yes 1984 Reagan (R) 86.1 Yes Hart (D) 37.3 No 1988 Bush (R) 37.6 Yes Dukakis (D) 35.7 Yes 1992 Bush* (R) 53.0 Yes Tsongas (D) 33.2 No 1996 Clinton* (D) 83.9 Yes Buchanan (R) 27.2 No 2000 Gore (D) 52.0 Yes McCain (R) 49.0 No

Note: Entries are the winners of all primaries from 1912 to 1948, and the winners of the first primary (New Hampshire) from 1952 onwards. An asterisk identifies an incumbent president. Sources: Apple (2000); Congressional Quarterly (1994, 489-560); Pomper (1997, 46, 60).

46 PS March 2001



the vote in the immediately previous election and negatively correlated with the vote two elections before, with both effects being roughly equal in abso- lute size.2 The positive sign for the VOTEt-1 parameter indicates that a party winning a presidential election can expect to hold on to much of its above-equilibrium vote portion in the immediately following election. At the same time, the winning party must reckon with a reversal in elections after that, as signified by the negative sign for VOTEt-2. In other words, parties seem to incur a "third-term penalty" (Abramowitz 1988). That is not to say that a party is doomed to lose after eight years in office, only to return to power after two terms in opposition. Since the effects of the immediate past vote and the vote two elections before are nearly equivalent, the incumbent party may indeed be able to hold on for a third term, but rarely longer than that. The average length of the presidential vote staying above (or below) the 50% mark is 2.5 terms.3

The forecast model I used for the 2000 election incorporated the autoregressive dynamic and primary performance of the in-party candidate, which was entered into the equation as a simple binary variable, scored positive or negative (according to the definition provided above). Preliminary tests showed that primary performance of the out-party's candidate did not significantly affect the November vote. Note that the VOTE variables refer to the Republi- can percentage of the major-party vote, regardless of whether or not a Republican held the presidency. That called for inverting the primary- performance variable for election years with Democrats in control of the White House.

As can be seen below, the primary performance of the incumbent-party candidate (PP) proves to have an enormous and hugely significant effect; worth more than five percentage points for the incumbent-party candidate if he were the primary winner.

VOTEt = 52.5 +5.40

PPt +.37VOTEr1 -.41VOTE-2 (5.2) (.78) (.11) (.09)

Adj.R2 = .860 SER = 3.18 N = 22 (1912-96)

At the same time, the model estimates also confirm the cyclical autoregressive dynamic of presidential elections. Had the autoregressive failed to show signs of life in the model featuring primary performance, analysts would have had to conclude that primary performance was at the heart of the autoregressive

dynamic. That certainly is not the case. Instead, evidence shows that the cyclical pattern of the vote is largely independent of primary performance in a given election year. Considering the incumbent-party's primary performance and the cyclical dynamic of the

TABLE 2 Primary Performance and General Elections, 1912-1996

Incumbent-Party Candidate Out-Party Candidate

Primary Performance Primary Performance

General Election Positive Negative Positive Negative Winner 13 0 7 2 Loser 1 8 2 11

Source: Table 1 and election returns.

vote led me to forecast that, come Election Day, Al Gore would outpoll George W. Bush 55 to 45% among two-party voters.

How much confidence did I have in this forecast? To be honest, contrary opinion poll results throughout the

TABLE 3 Opinion of Bill Clinton and Vote Choice, General Election and New Hampshire Primary, 2000

Opinion of General Election New Hampshire Primary B ill C lin to n------- ---------- ------- as President Vote for Vote for Vote for Vote for and Person All Gore Bush All Gore Bradley

Approve/ 35% 85% 12% 41% 68% 32% Like

Approve/ 20% 63% 33% 37% 47% 52% Dislike

Disapprove/ 1% - - 1% - -

Like Disapprove/ 39% 7% 89% 16% 24% 74%

Dislike

Source: Voter News Service election-day exit polls; 13,130 voters in general election; 1,730 in (Democratic) New Hampshire primary. The questions about Clinton read: "Do you approve or disapprove of the way Bill Clinton is han- dling his job as president?" and "Is your opinion of Bill Clinton as a person favorable or unfavorable?" Voters without opinions or vote choices are not shown; hence percentages do not add to 100.

year until Labor Day made me somewhat uneasy (Berke and Elder 2000; Newport 2000). But the past record of the model reassured me. And, with a standard error of three points, the forecast of 55% for Gore appeared safe. While the 95% confidence interval (roughly 2 standard errors) meant I could not entirely rule out a dead heat or a narrow Bush victory, I was at least sure that the chances of a Bush victory in the popular vote were a paltry one in ten. There is no denying that the error of the 2000 forecast is dismayingly large. What went wrong?

PSOnline www.apsanet.org 47



What Went Wrong? There are too many suspects to examine and too little

data, at this point, to allow me to reach a verdict at this time. I can, however, rule out one suspect: the out-party's performance. Tests for its influence on the vote during the 1912-96 period came up empty. I have already dealt with another suspect by including it in the model: the fatigue factor. The cyclical dynamic variables in the model capture in-party fatigue more than adequately. Barring special advantages, the incumbent party could hope for no better than an even split in 2000. So what happened to the edge Gore was predicted to reap from his primary success? Incumbent primary performance is considered a leading indicator of the vote in the general election because it reflects public approval of the sitting president. The case is simple if and when that sitting president himself is running in the election. But what if someone else in his party reaches for that mantle?

In the 2000 election, as in 1960, a sitting vice president sought to succeed a two-term president with high approval ratings. Like Nixon, Gore struggled to carve his own profile while taking advantage of the good will left behind by his predecessor. In both cases, the incumbent-party candidates were unable to convert that good will into a comfortable victory. Why? Perhaps it is to be expected that one man should not be judged on the record of another one, even if they are of the same party and served in the same administration. Perhaps the forecasts gener- ated with my model would have been right had Eisenhower and Clinton been allowed to run for third terms.

As for the 1960 contest, it was easy to see that Eisenhower's approval rating would not transfer to any successor. He was popular for largely personal reasons, not for policy achievements during his tenure. In fact, neither peace nor prosperity seemed assured in the fall of 1960. In 2000, however, Clinton's approval ratings rested strongly on a policy foundation. Yet, again the vice president ran into obstacles trying to parlay that approval into a vote for him. Why this time? Consider the possi- bility that Bill Clinton-the "primary color" president- was a mixed blessing for Al Gore. The vice president certainly kept a pointed distance from the president during the campaign and allowed the contest to be framed as one over more versus less government.

The election-night exit poll offers some suggestive evidence (Toner and Elder 2000). While most Americans approved of Bill Clinton as president, most voters (not the same group) viewed him unfavorably as a person. Ill feelings about Clinton personally cut Al Gore's support among those approving of Clinton's job as president from 85 to 63%, as Table 3 shows. That may have been enough to deprive Gore of a comfortable victory.

Could it be that Al Gore paid for Bill Clinton's sins? The mixed blessing of Clinton's legacy was already evident in the first and foremost primary election. Al Gore just barely edged his challenger Bill Bradley in New Hampshire (52 to 47%), following a back-and-forth struggle in the preceding months. It was perhaps the close call for Gore in the New Hampshire primary, not his victory, that was the true omen for the general election.

Notes *1 am grateful to numerous people for their comments, curiosity, and

questions about this forecasting exercise, especially Bob Kaiser, Mike Kagay, Dan Merkle, James Warren, Adam Clymer, David Miller, Alan Elsner, and Katherine Reynolds. The forecast was presented at the Annual Meeting of the American Political Science Association. Thanks go to Jim Garand for setting up the panel.

1. Until 1952, no single state with a primary could count on playing the lead role in the presidential-selection drama. That changed when New Hampshire officials decided to put presidential candidates rather than convention delegates on the state's primary ballots. That switch "gave presidential hopefuls an opportunity to demonstrate early strength" (Buell 2000, 93) and propelled New Hampshire into the most coveted spot of the primary season (Adams 1987). Winning in New Hampshire, however small and unrepresentative the state's electorate might be, boosts a presidential candidate's hope for a nomination

victory in a way no other primary win can match. 2. The parameter estimates of the AR(2) model are such that the

equilibrium (stationarity) conditions are met:

(1) *1+0 1<1; (2) ?;-+< 1; (3) -1 < )2 < 1.

3. Considering just this cyclical dynamic allowed me to very accurately predict Clinton's reelection and share of the national two- party popular vote in 1996 (Norpoth 1995). For 2000, however, sole consideration of the cyclical dynamic showed that a Republican candidate could expect 50.2% of the two-party vote, while a third- term seeking Democratic candidate could only expect 49.8% of the popular vote. That forecast hit the ultimate result closer than that of any other model. But it could have been far off the mark, given the forecast's large standard error. That it came so close to the ultimate result nonetheless offers some food for thought on how to read the outcome of the 2000 election.

References Abramowitz, Alan I. 1988. "An Improved Model for Predicting

Presidential Election Outcomes." PS 21(December):843-47. Adams, William C. 1987. "As New Hampshire Goes..." In Media and

Momentum, ed. G. Orren and N. Polsby. Chatham, NJ: Chatham House.

Apple Jr., RW. 2000. "A Beginning, Not an End." The New York Times, February 2, Al.

Atkeson, Lonna Rae. 1998. "Divisive Primaries and General Election Outcomes: Another Look at Presidential Campaigns." American Journal of Political Science 42:256-71.

Berke, Richard L, and Janet Elder. 2000. "Poll Finds Voters Relying on Issues to Shape the Race." The New York Times, July 25, Al.

Buell, Emmett H. 2000. "The Changing Face of the New Hampshire Primary." In In Pursuit of the White House 2000, ed. W. Mayer. New York: Chatham House.

Campbell, James E., and James C. Garand, eds. 2000. Before the Vote. Thousands Oaks, CA: Sage.

Congressional Quarterly. 1994. CQ Guide to U.S. Elections. 3rd ed.

Washington, DC: CQ Press. Lewis-Beck, Michael S., and Tom W. Rice. 1992. Forecasting

Elections. Washington, DC: CQ Press. Midlarsky, Manus I. 1984. "Political Stability of Two-Party and

Multiparty Systems: Probabilistic Bases for the Comparison of Party Systems." American Political Science Review 78(September):929-51.

Newport, Frank. 2000. "Bush Expands Lead Over Gore" <www.gallup.com/poll/releases/pr000627.asp>. Gallup Poll, June 27. Accessed: January 27, 2001.

Norpoth, Helmut. 1995. "Is Clinton Doomed? An Early Forecast for 1996." PS: Political Science and Politics 28(June):201-07.

.2000. "Primary Colors: A Forecast of the 2000 Presidential Election." Presented at the Annual Meeting of the American Political Science Association, Washington, DC.

Pomper, Gerald, ed. 1997. The Election of 1996. Chatham, NJ: Chatham House

Toner, Robin, and Janet Elder. 2000. "An Electorate Largely Split Reflects a Race So Very Tight." The New York Times, November 8, B1.

48 PS March 2001



Documents

Primary Colors: A Mixed Blessing for Al Gore